Ethereum

[jl777]

https://bitcointalk.org/index.php?topic=1394020.msg14161331#msg14161331

"I think it's more likely that Bitcoin Core would eventually move to a custom data structure than to another "database"

So at least the existing devs are listening to the iguana. Unfortunately the doublespeak persists: "Bitcoin Core does not store the blockchain in a database (or leveldb) and never has. The blockchain is stored in pre-allocated append only files on the disk as packed raw blocks in the same format they're sent across the network"

Somehow the fact that the blockchain is just stored as raw data and there is a DB on top of it, means the core doesnt use DB? I guess technically the blockchain data isnt in a DB, just all the data structures needed to do anything with it. not sure why he thinks that is such an important distinction in the context of removing DB dependence

 

[YarkoL]

I claim it is possible to create ANY Dapp on top of bitcoin that ETH can create and the result will be more efficient and lightweight and even mass market possible.

How about loops, e.g. iterating through a key-value mapping?
Or do you have in mind a scenario where the app is implemented solely by metadata (and not by stack script), Mastercoin-style?

 

[jl777]

I think the approach of putting everything in the blockchain is ... not efficient

why should transient data have to go into a blockchain at all? By proper partitioning of the data between temporary, permanent, money, non-money, etc. you can reduce the amount that needs to go into the blockchain by orders of magnitude

It seems overkill to use DB for permanent data of fixed format. Likewise it is overkill to use blockchain for things that dont need global consensus. Have people forgotten about hierarchical methods?

Separately, why would there be a loop for key-value mapping? You can easily get constant time performance using a hash table
[doublepost=1457708887][/doublepost]for almost turing functionality without problems regarding partioning, just use FSM with data from confirmed blocks

 

[Matthew Light]

Ethereum already has support for putting other tokens on it such as Bitcoin. That was part of the design from day 1. I don't know the details,but I'm sure it scales well.
[doublepost=1457709288][/doublepost]

can someone explain how a 20GB/yr blockchain before any major usage is anything that can be run in a decentralized way?

I bought a 5TB last year for $120. That's 250 years of the last year of Ethereum transactions.
[doublepost=1457709353][/doublepost]

I think the approach of putting everything in the blockchain is ... not efficient

why should transient data have to go into a blockchain at all? By proper partitioning of the data between temporary, permanent, money, non-money, etc. you can reduce the amount that needs to go into the blockchain by orders of magnitude

It seems overkill to use DB for permanent data of fixed format. Likewise it is overkill to use blockchain for things that dont need global consensus. Have people forgotten about hierarchical methods?

Agreed. Blockchain storage in Ethereum is relatively expensive, as it should be.

 

[jl777]

well chrome apps only allow 10GB total and getting bitcoin squeezed in there is hard enough.

centralizing everything on a single blockchain that will eventually be KYC'ed does not seem like a good idea

If ETH scales up its usage, the blockchain growth might need TB/yr. It is not a matter of HDD space, but of bandwidth. bitcoin will be able to scale at least 10x more than ETH, maybe 100x. I dont see ETH going beyond financial contracts and for that it is great, but what sort of bandwidth is needed just to keep up with its TB/yr blockchain? Let alone to bootstrap it.

"hold on, I just need to catch up on the last 292 GB of blockchain before I can do the tx"

ETH is like a battlefield tank. It is great for things that need a tank, but it makes bitcoin look like a small footprint.

 

[Matthew Light]

If ETH scales up its usage, the blockchain growth might need TB/yr.

ETH monetary transactions take up less space and bandwitch than BTC ones (Accounts are more efficient than UTXOs).

centralizing everything on a single blockchain that will eventually be KYC'ed does not seem like a good idea

I'm curious what you mean by this? Why would ETH and BTC be any different with respect to "KYC"?

 

[jl777]

who will be able to run nodes that can handle the TB/yr blockchains (not HDD space, but bandwidth)?

this leads to conclusion that SPV nodes are the only way ETH can achieve any semblance of large scale use. So, there will be a limited set of nodes that all other nodes connect to. And with such a network structure, how hard would it be to popup an alert saying that you need to do KYC to keep using the servers?

Since it would take a massive amount of bandwidth to bootstrap the full chain by then, what choice will people have? It will start with asking for email so they can be contacted in urgent situations and it will never be divulged. Then the next tidbit of information. Probably the process will take years, "boil the frog by gradually increasing the water temp" approach

Of course you can trust that a corporate backed entity will never, ever, do any such thing, in spite of pressure, penalties, threats of jail time, etc that would be applied to the hosts. But wait! These servers will most likely be run by the big corporations, so they wont even fight being regulated. After all that is the way it has always been

My prediction is based from my point of view as an implementor. I just dont see how a non-SPV ETH is even possible, if it succeeds. If it doesnt, then it is a moot point.

Whoever controls the network, will control the network. I dont think this point can be argued with. And once you have control of a valuable network, then you maximize its value, from your perspective. Since avoiding costly battles with fiat govts is logical, that is what they must do.
[doublepost=1457714164][/doublepost]I am able to write a scalable bitcoin core, but I am unable to even see how it is possible to do for ETH

 

[YarkoL]

Separately, why would there be a loop for key-value mapping? You can easily get constant time performance using a hash table

I was just trying to think up some example, but I'm sure there could be better ones. I'm by no means expert, just getting re-acquainted with Solidity, as it is now much better documented than it was last summer.
I think you are right about the scaling issues, validation is going to be restricted to a select few with a lot of moolah.
Still, Eth is a damn fun to play with!

 

[Matthew Light]

Since it would take a massive amount of bandwidth to bootstrap the full chain by then

The chain is pruneable.

this leads to conclusion that SPV nodes are the only way ETH can achieve any semblance of large scale use. So, there will be a limited set of nodes that all other nodes connect to. And with such a network structure, how hard would it be to popup an alert saying that you need to do KYC to keep using the servers?

If Ethereum takes off, there will be at least hundreds of thousands of fully-validating nodes (now granted, with sharding they won't all contain the entire blockchain which will be distributed across shards).

 

[jl777]

if it is pruned, then archival nodes will need to be trusted. Guess who will be running the archive nodes?

hundreds of thousands? Really. Amazing how many charitable people there are all just hosting a multi-TB/yr blockchain with hundreds of dollars per month bandwidth cost. Why not claim millions or tens of millions of people would? Even better, a billion people would. See I can make up numbers just as good as you. However, historically speaking when people have to keep paying for something that doesnt get them anything in return, they stop. And since the corporate run servers provide SPV service, there wont be any incentive for the masses to run a node. Other than altruists and purists. There are 100,000+ of those around? What basis do you have to support this?

And with sharding, how exactly will data from different shards get synchronized? I know! via the massive nodes that are run by? Guess who, yes, the same ones running the archive nodes. And what happens if there is a dapp that accesses data across all the shards?

So you just support my position that technical forces combined with economic and bandwidth constraints will force ETH chain to become centralized. And once only large corporations are having the full blockchain/shards/archives, how hard is it for pressure to be applied for KYC conformance

 

[Matthew Light]

hundreds of thousands? Really. Amazing how many charitable people there are all just hosting a multi-TB/yr blockchain with hundreds of dollars per month bandwidth cost.

It's not about charity.

It is very likely that Ethereum will be the foundation of our global financial system. So people vested in control of significant sums of Ethereum / money will certainly run their own nodes out of a desire to ensure that their funds are protected.

And with sharding, how exactly will data from different shards get synchronized?

Different shards don't get synchronized. Each shard "flavor" has its own unique blockchain. That way everyone doesn't need a copy of all the data, which drastically reduces bandwidth requirements.

 

[jl777]

You convinced me. I have sold everything I own and bought ETH with it

 

[Matthew Light]

You convinced me. I have sold everything I own and bought ETH with it

If you had only done that two months ago...

 

[YarkoL]

It's not about charity.
Different shards don't get synchronized. Each shard "flavor" has its own unique blockchain. That way everyone doesn't need a copy of all the data, which drastically reduces bandwidth requirements.

Is that really the plan? I find it hard to understand what's going on. Bitcoin is difficult enough as it is, and reading a paragraph of Vitalik is like some arcane treatise in number theory. On the other forum Anonymint gives the impression that he's got it all figured out and I'm always unsure if he's a genius or a maniac... All I know that I must be pretty stupid 'coz I'm not gettin it

So what determines these shards, is there a constant number of them? Is there some asynchronous calling protocol that enables them to message each other?

 

[Matthew Light]

And we are at 1 Billion market cap now!

 

[8up]

Everything moves in waves.

Even when ETH runs to 0.05/$25 (which is easily possible), I think we will see a back flow into FIAT and BTC (also maybe LTC) very soon.

Risk/reward is not the same as in Q3/Q4 2015 anymore. And the big migration (if it really should take place) will have phases of backlash.

 

[Matthew Light]

Even when ETH runs to 0.05/$25 (which is easily possible), I think we will see a back flow into FIAT and BTC (also maybe LTC) very soon.

Depends on whether this is a migration of value from one blockchain to another, or just a bubble.

If it is a genuine migration, then the best strategy for BTC holders is to hurry up and do it before BTC market cap deflates by 50-75%.

 

[_mr_e]

Except a single pool has > 50% of the Ether hashpower now so no thank you: https://etherchain.org/statistics/miners
[doublepost=1457966544][/doublepost]I also bet part of the reason the price has gone up so high is big corporations like Microsoft buying up huge chunks so they can forever own the POS mining once the switch happens. This will one day be a huge problem for Ether I fear.

PoW keeps miners constantly upgrading and fighting it out with eachother whereas with Ether all Microsoft has to do is sit on their coins for all eternity and will have a huge say in the direction of the protocol moving forward.

 

[Bloomie]

Microsoft didn't buy any ETH. They just support hosting Ethereum development environments on Azure.

 

[_mr_e]

Microsoft is just an example. It's impossible to know who has bought what and how long they'd be able to hold mining hostage with that stash once it goes POS.